agents-md-best-practices

Refactor AGENTS.md instructions to follow progressive disclosure and AGENTS.md best practices. Use when asked to audit, reorganize, or split AGENTS.md into a minimal root file plus linked docs; includes contradiction checks, essential extraction, grouping, and deletion flags.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "agents-md-best-practices" with this command: npx skills add heykvnzhao/dotfiles/heykvnzhao-dotfiles-agents-md-best-practices

Agents MD Best Practices

Overview

Refactor an AGENTS.md into a minimal root file with links to focused docs, following progressive disclosure. Produce contradictions to resolve, a root essentials set, grouped files, a suggested docs/ structure, and deletion flags.

Workflow

0) Gather inputs

  • Ask which AGENTS.md to refactor (path or pasted content).
  • Ask whether to apply file changes or draft only.
  • Ask where to place new docs/ (default: docs/agents/).

1) Parse and normalize

  • Read the AGENTS.md content.
  • Split into individual instructions (one per line/bullet/paragraph).
  • Keep original wording; annotate with a short label so you can reference items precisely.

2) Find contradictions

  • Identify instructions that conflict (opposing directives, incompatible tools, or contradictory constraints).
  • Present each contradiction as a pair (or set) and ask which to keep.
  • Pause before restructuring until contradictions are resolved.

3) Identify essentials for root AGENTS.md

Keep only what belongs in the root file:

  • One-sentence project description.
  • Package manager if not npm.
  • Non-standard build/typecheck commands.
  • Anything truly relevant to every single task.

Ask if the user wants any other item in root before finalizing.

4) Group the rest into categories

  • Cluster remaining instructions into logical, named categories (e.g., TypeScript, Testing, API design, Git workflow, UI, Infra).
  • Create a separate markdown file per category.
  • Keep categories actionable and scoped; avoid mixed topics.

5) Create the file structure

Provide:

  • A minimal root AGENTS.md with markdown links to the category files.
  • Each category file with its relevant instructions.
  • A suggested docs/ folder structure.

Use the template in references/output-template.md when helpful.

6) Flag for deletion

Identify instructions that are:

  • Redundant (agent already knows this).
  • Too vague to be actionable.
  • Overly obvious (e.g., “write clean code”).

List these separately with a short reason and ask for confirmation before removing.

Output requirements

  • If no contradictions, explicitly say “No contradictions found.”
  • Use clear section headings in the response: Contradictions, Root AGENTS.md, New Files, Suggested docs/ Structure, Flags for Deletion.
  • Provide file contents in fenced code blocks with filenames.

Guardrails

  • Do not delete or rewrite instructions without user confirmation when conflicts exist or items are flagged.
  • Do not expand scope beyond the AGENTS.md content unless the user asks.

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

General

git-best-practices

No summary provided by upstream source.

Repository SourceNeeds Review
General

coding-best-practices

No summary provided by upstream source.

Repository SourceNeeds Review
General

git-commit-plan

No summary provided by upstream source.

Repository SourceNeeds Review
General

web-ui-best-practices

No summary provided by upstream source.

Repository SourceNeeds Review